Interesting thread... In my current setup, I do not have a single indicator. Not even volume all tho I recognize and read it ASAP due to the immediate effect it causes. I guess if I were to have an indicator, it would be any leading indicator although many indicators can be made leading using some relative change calculations. I don't like them because I am lazy (with a pencil that is) and furthurmore it only measures what has already happened. I let the market do ALL the calculations. It is interesting how the context for (IMO) the only existing leading indicator has an objective context and the information on it changes more rapidly than a single tick. In any event, it is a single direct indicator, it shows you exactly what hasn't shown up on your chart yet. Your price chart is only showing you what already happened. Evidently, even tho it exists, chart programs have not caught up (in the continual money making sense) as of yet. In any event, I had realize my own orientation, and then worked out what it means to know what won't happen since I profit exactly from what didn't happen. This type of indicator scenario is non numeric... Then the idea of convential indicators becomes silly. Even scarier is that there is a force at play that results in the action (or inaction as the case may be for non winners) displayed on your leading indicator. The force (not measured in newtons) is at an exact Normal to the plane of your chart. This is probably way OT I presume. Alec: What is a DOM? Trader: "Fill in your Q!"
First you say: "indicators cannot tell you something that is not already on the chart" Then you say: "they may be helpful to filter the noise in the price data and to make it easier for the eye to spot trends which is the primary goal of a speculator" There is WAY too much confusion in your statements. Yes, simple mainstream indicators will tell you something you may already see in a price chart. However, more advanced indicators will tell you things you may never be able to see visually, usually by considering a lot more market data than one person can memorize, for the purposes of automatic adaptation or calibration.
Actually, good indicators convey more accurate information than the raw data unless you assume we are all savants that can apply numerically complex models in our heads. That is why we use indicators. Indicators make make patterns and opportunities more obvious so our brains can focus on the big picture. There need not be a distinction between indicators and experience. Indicators are based on experience! There are two examples of this. First you can have a trader use their experience to define and adjust an indicator. Secondly you can have an indicator that uses experience automatically through backtesting, either to aid an ATS or discretionary trading. The best indicators I've ever seen, profitability-wise, rely on experience specifically. You probably know about such things through your A.I background. Yes these beasts also have the form "Indicator = F( Time, Price, Volume);" and they truly are indicators. They can convey significantly more information than the raw data through advanced filtering, adaptation or calibration, which are all really the same thing, namely acting on experience. You can also add microstructure, volatility and cross-market inputs too, though price volume and time are far more useful and reliable, in my experience.
Definitely. Backtesting is also essential. How many different indicator settings can one trader possibly watch in real time simultaneously? A lot less than one can consider through backtesting. Anyone not backtesting their ideas or indicators is giving up a huge advantage.
You mean depth of market? Yes historical DOM is available. I have several indicators that make use of DOM structure and dynamics. Some need to go deep (>=10 levels), some that only need inside depth. One can also build indicators on order transaction data (adds, changes, deletes, executions, busts), or any other market information. Order transaction data is the real gold mine. You see everything there.
Hi Thunderdog, Of course you use arithmetic on price data. (Don't feel bad about not sharing! Nobody shares something that really works). Coming back to indicators, this is all semantics. In fact I don't like the name indicators at all. However, it all comes to the same thing, once you are using algorithmic procedures to arrive at your trading decisions, you are by necessity dealing with 'decision variables', very complex or very trivial perhaps. It is not important how you label such variables, x, y47, filter23_output or MACD: it comes to exactly the same thing. Some people don't have your talents and have to do with 'canned' variables popularly known as indicators. I don't exclude the possibility for somebody to artfully distill a profitable method out of these canned indicators (I'm not sure that's the right term to use). They are in essence no different from yours. Those ridiculous unending threads about trend or not is a pure case in point. Obviously people want to say something about applying arithmetic to price. (Nobody did, unless I missed it!) If you fail to recognize this, you can argue a long time about nothing. Admitting you use crystal balls of rabbit's feet for trend finding doesn't look right at ET! Be good, nononsense
IMO, the market is always oscillating up and down in cycles. An indicator that is in tune with the cycles is extremely useful. The macd is useful if it is in tune with the cycles of the market. Jack has suggested the 5,13,6 macd. IMO, the histogram and 6 ma is of no use to really see what is going on and actually quite confusing. It is more useful to use a 5,13,1 macd or a 5 ema and 13 ema on ones chart. Once you see the oscillations of the market, you should always know where you are. Then, volume is useful for indicating peaks and troughs of cycles and the strength of the new half cycle. This is just the opinion of a 3 year newb